# MYCIN **Domain:** Medical AI / Expert Systems / Governance **Doc Type:** Canonical System Node **Maturity:** Foundational ## Definition **MYCIN** was a Stanford rule-based expert system developed in the 1970s, principally through Edward Shortliffe's doctoral work, to advise on diagnosis and antimicrobial treatment for serious bacterial infections. ## Architecture and Evaluation MYCIN used production rules, certainty factors and an interactive consultation. Historical blinded evaluations found its recommendations competitive with specialists under constrained test conditions. Those results should not be confused with a prospective clinical trial or routine deployment. MYCIN did not enter ordinary clinical use. The barriers included integration, responsibility, explanation, maintenance and the institutional status of computer-generated medical advice—not a single simple “liability” cause. ## Evolutionary Nexus [[articles/The Evolutionary Roots of Silicon Valley|The Evolutionary Roots of Silicon Valley]] pairs MYCIN with DENDRAL to show that two foundational expert systems were biological or medical. It also turns MYCIN into an early case of capability outrunning custody. ## Rights Context A system can outperform some clinicians on a bounded task without possessing authority to treat a patient. Deployment requires accountable operators, patient consent, audit, update control and recourse. ## Sources / Provenance - Stanford Biomedical Data Science alumni and program history: https://dbds.stanford.edu/alumni/ and https://dbds.stanford.edu/biomedical-data-science-graduate-program-overview/ - Yu et al., “Antimicrobial Selection by a Computer,” _JAMA_ 242 (1979), record: https://profiles.stanford.edu/stanley-cohen ## See Also [[wiki/Edward Shortliffe|Edward Shortliffe]], [[wiki/Expert System|Expert System]], [[wiki/Clinical Decision Support|Clinical Decision Support]], [[wiki/Moral Accountability Layer|Moral Accountability Layer]]